Analysis of Malware Detection in Downloaded Files Based on CNN-SVM-RBF Approach

Abstract

It is possible for malicious coders to take advantage of programming errors committed by the developers themselves. These blunders can be the result of sloppy coding or logical flaws in the original design. Low-level languages like C and C++ make it possible to use pointers, which allow for arbitrary memory access within a program's address space. Unwary programmers' misuse of these features can result in illegal memory reads and writes. Malicious programmers use these unapproved reads and writes to cause even more devastation in the system. The proposed procedure begins with three stages: preprocessing, feature extraction, and model training. Normalization and regularization of data are part of the preprocessing phase. In order to pick features, extract N-gram features, and utilize Windows API functions. After features have been selected, the models are trained with CNN-RBF-SVM. The proposed method outperforms the two most used alternatives, RBF-SVM and CNN.